Figure 5.

Design effective experiment for a linear system. (Example 4) Reconstruction results based on different experiments: Exp#1 represents the original experimental dataset which has limitation to reduce coherence; Exp#2 represents non-effective experimental dataset; and Exp#3 represents the effective experimental dataset (A) the sensing matrix and reconstruction result (B) coherence distribution comparison. This example demonstrates that if there is a limitation to reduce coherence in the original dataset, we can also design a more effective experiments based on the coherence distribution. By revealing deficiency in the data, the proposed method can help reduce the cost of experiments.